config-scan

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION] (LOW): The skill instructions include shell commands for file discovery and content analysis.
  • Evidence: The skill uses git ls-files | grep and various regex patterns to analyze file content.
  • Scope: This is restricted to local, read-only operations on the user's repository to identify misconfigured files.
  • [PROMPT_INJECTION] (LOW): The skill is susceptible to Indirect Prompt Injection because it processes untrusted external data (configuration files) to generate security reports.
  • Ingestion points: Processes .env, Dockerfile, docker-compose.yml, Kubernetes YAML manifests, and Terraform (.tf) files.
  • Boundary markers: None. The skill interprets the content of these files directly using regex patterns.
  • Capability inventory: Read local files, execute basic shell commands (git, grep), and provide evaluative feedback to the user.
  • Sanitization: No explicit sanitization or instruction-filtering is mentioned for the content being scanned.
  • Severity Justification: Classified as LOW (Tier: Internal influence only) because the skill's output primarily influences the agent's reasoning/reporting rather than triggering high-privilege write operations or network exfiltration.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 08:53 AM